Merge pull request #3 from getwilds/add-gpu-test

Adding GPU test WDL

.github/workflows/womtools-validate.yml

@@ -29,6 +29,7 @@ jobs:
         run: |
           java -jar womtool-86.jar validate helloHostname/helloHostname.wdl
           java -jar womtool-86.jar validate helloDockerHostname/helloDockerHostname.wdl
+          java -jar womtool-86.jar validate gpuMatrixMult/gpuMatrixMult.wdl
           java -jar womtool-86.jar validate parseBatchFile/parseBatchFile.wdl -i parseBatchFile/parseBatchFile-inputs.json
           java -jar womtool-86.jar validate variantCalling/variantCalling.wdl -i variantCalling/variantCalling-inputs.json
 

gpuMatrixMult/gpuMatrixMult.wdl

@@ -0,0 +1,57 @@
+version 1.0
+## This is a test workflow that performs simple matrix 
+## multiplication using a single GPU. 
+
+#### WORKFLOW DEFINITION
+
+workflow GpuMatrixMult {
+  call Hello {
+  }
+
+  output {
+    File stdout = Hello.response
+  }
+
+  parameter_meta {
+    stdout: "printed results of the gpu script in question"
+  }
+}
+
+#### TASK DEFINITIONS
+
+task Hello {
+  command <<<
+    set -e
+    echo 'hello there from the WILDS' 
+
+    python3 <<CODE
+    import tensorflow as tf
+    # Creates a constant tensor from a tensor-like object.
+    a = tf.constant([1.0, 2.0, 3.0, 4.0, 5.0, 6.0], shape=[2, 3], name='a')
+    b = tf.constant([1.0, 2.0, 3.0, 4.0, 5.0, 6.0], shape=[3, 2], name='b')
+    # Multiplies two tensors element-wise.
+    c = tf.matmul(a, b)
+    # Prints the result.
+    print(c)
+    # Check if GPU is available
+    print("Num GPUs Available: ", len(tf.config.experimental.list_physical_devices('GPU')))
+    CODE
+
+    echo 'bye for now'
+  >>>
+
+  output {
+    File response = stdout()
+  }
+
+  runtime {
+    docker: 'tensorflow/tensorflow:latest-gpu'
+    # modules: "TensorFlow/2.11.0-foss-2022a-CUDA-11.7.0"
+    gpus: '1'
+  }
+
+  parameter_meta {
+    response: "printed results of the gpu script in question"
+  }
+}
+